Cellulose Synthase-Like D1 is integral to normal cell division, expansion, and leaf development in maize.

The Cellulose Synthase-Like D (CslD) genes have important, although still poorly defined, roles in cell wall formation. Here, we show an unexpected involvement of CslD1 from maize (Zea mays) in cell division. Both division and expansion were altered in the narrow-organ and warty phenotypes of the csld1 mutants. Leaf width was reduced by 35%, due mainly to a 47% drop in the number of cell files across the blade. Width of other organs was also proportionally reduced. In leaf epidermis, the deficiency in lateral divisions was only partially compensated by a modest, uniform increase in cell width. Localized clusters of misdivided epidermal cells also led to the formation of warty lesions, with cell clusters bulging from the epidermal layer, and some cells expanding to volumes 75-fold greater than normal. The decreased cell divisions and localized epidermal expansions were not associated with detectable changes in the cell wall composition of csld1 leaf blades or epidermal peels, yet a greater abundance of thin, dense walls was indicated by high-resolution x-ray tomography of stems. Cell-level defects leading to wart formation were traced to sites of active cell division and expansion at the bases of leaf blades, where cytokinesis and cross-wall formation were disrupted. Flow cytometry confirmed a greater frequency of polyploid cells in basal zones of leaf blades, consistent with the disruption of cytokinesis and/or the cell cycle in csld1 mutants. Collectively, these data indicate a previously unrecognized role for CSLD activity in plant cell division, especially during early phases of cross-wall formation.


INTRODUCTION
The ancient, highly-conserved family of Cellulose Synthase-Like D (CSLD) proteins are required for cell growth and development, yet their biochemical and cellular functions are only now emerging (Richmond and Somerville, 2000;2001;Favery et al., 2001;Wang et al., 2001;Bernal et al., 2007;2008;Yin et al., 2009;Park et al., 2011;Yin et al., 2011). CSLDs belong to one of ten distinct subfamilies in the Cellulose Synthase superfamily, defined by amino acid sequence similarity to Cellulose Synthase (CESA) (Richmond and Somerville, 2000;Hazen et al., 2002;Farrokhi et al., 2006;Penning et al., 2009;Fincher, 2009). All members of this superfamily share predicted function based on sequence identity as membrane-bound, processive glycosyltransferases that synthesize beta-linked glycan polymers, such as those of cell wall polysaccharides (Richmond and Somerville, 2000;2001). Known products range from cellulose to hemicellulose backbones, and may include additional beta-linked glycan chains (Arioli et al., 1998;Dhugga et al., 2004;Liepman et al., 2005;Burton et al., 2006;Cocuron et al., 2007;Doblin et al., 2009). The CSLDs remain poorly understood despite their importance in cell development and evidence for their evolution in plant lineages extending back to non-vascular land plants and possibly before (Roberts and Roberts, 2007).
Of the cellulose synthase-like genes, CslDs are the most closely related to the CesAs themselves, leading to early suggestions that CSLDs may also function as cellulose synthases (Doblin et al., 2001). The CSLD proteins also share the greatest amino acid sequence identity (40-50%) with CESAs and are similar or slightly larger in size (Richmond and Somerville, 2001). Other than CESAs, CSLDs are the only members of the superfamily that have the aminoterminal, Zn-finger-like domain thought to function in protein-protein interactions, possibly mediating formation of complexes, or protein turnover (Richmond and Somerville, 2000; Kurek 7 2011). Polarized plasma membrane localization of CSLD3 supports a role in polysaccharide deposition at the growing tip of Arabidopsis root hair cells (Park et al., 2011).
Thus far, the majority of genetic evidence has implicated CSLDs in polarized cell expansion associated with polar tip growth typical of pollen tubes and root hairs (Bernal et al., 2008Favery et al., 2001Wang et al., 2001;Kim et al., 2007;Bernal et al., 2008;Penning et al., 2009), and the intrusive tip growth typical of xylem and other sclerenchyma fibers (Samuga and Joshi, 2004). In the moss, Physcomitrella patens, where tip growth predominates in caulonema cells, CslD genes comprise 46% of all expressed sequence tags from the CESA superfamily, including all CesAs and CSLs (Roberts and Bushoven, 2007). In vascular plants, root hairs and pollen tubes provide classic models for cellular tip growth (Hepler et al., 2001;Cole and Fowler, 2006), but xylem and sclerenchyma fibers also elongate by a form of intrusive tip growth (Mellerowicz et al., 2001;Samuga and Joshi, 2004). Expression of CslD2 in developing xylem of Populus is consistent with its proposed influence on xylem fiber length (Samuga and Joshi, 2004). However, the function of CSLDs in other aspects of cell growth, such as cell division, non-polar cell expansion, or differentiation, has not been explicitly studied to date. Some mutant phenotypes cannot be readily explained by polar growth defects (Bernal et al., 2007;Li et al., 2009;Yin et al., 2011), suggesting an expanded role for CSLDs in other cellular processes.
Here we present evidence for an unexpected, but integral role for CSLDs in plant cell division. In csld1 mutants of maize, defects in cell division were identified as the underlying basis for the narrow-organ morphology, as well as characteristic epidermal warts. Using reverse genetics tools in maize, we identified several mutants whose phenotype is caused by transposon insertion in CslD1. The csld1 phenotype was uniquely informative for dissecting the relationship between division and expansion. Through a detailed cellular analysis, we demonstrate that altered cell division in csld1 null mutants reduces cell number and is an underlying cause of the narrow-organ morphology. In-depth analysis of wall positioning, cell shape, nuclear size, cell ploidy, wall architecture, and alterations in cell number indicated that defective cell division was an early consequence of a CSLD1 deficiency. These data provide new insight into the function of CSLD proteins in plant growth, extending our understanding of their roles from tip-growth, to include cell division in most or all organs.

Phylogenetic Analyses Indicate Early Evolution and Conservation of Divergent Roles for CSLD Proteins
Phylogenetic analyses (Fig. 1A) identified three distinct clades in the CSLD subfamliy and found these to represent three phenotypic classes of csld mutants. These three major phenotypic groups include, for Arabidopsis, pollen tube defects in csld1 and csld4 mutants, (Bernal et al., 2008), root hair defects in csld2 and csld3 mutants (Favery et al., 2001;Wang et al., 2001;Bernal et al., 2008), and reduced plant size in csld5 mutants (Bernal et al., 2007). Thus far, mutants in related clades of CslD genes in rice and maize have yielded phenotypes similar to those of Arabidopsis. Mutants of rice CslD1 (Kim et al., 2007), for example, and its maize homolog, CslD5 (Penning et al., 2009), result in root hair-deficient phenotypes. Additionally, alterations in rice CslD4 (closest homolog of Zm-CslD1 and At-CslD5), confer a narrow leaf and dwarf1 (nd1) phenotype (Li et al., 2009;Hu et al., 2010;Wu et al., 2010). Similar functional roles are indicated by the reduced-growth phenotypes common to all three of these mutants (Fig. 1A, Bernal et al., 2007;Li et al., 2009;Hu et al., 2010;Wu et al., 2010).
Collectively, these data suggest conservation of specific developmental roles for individual CSLD proteins in plants, and indicate that these arose early in plant evolution.

An Allelic Series of csld1 Mutants in Maize Enabled Functional Analysis
Seven independent loss-of-function mutants for the maize CslD1 gene were identified in reverse genetic screens, including two from the UniformMu maize population (University of Florida) and five from the Trait Utilities for Screening of Corn (TUSC) lines (Meeley and Briggs, 1995;McCarty and Meeley, 2009) (Pioneer Hi-Bred Int.) (Fig. 1B). The two UniformMu alleles, csld1-1 and csld1-2, were examined in the greatest depth because of their uniform genetic background (McCarty et al., 2005). Both of these were null for detectable expression of CslD1 mRNA by qPCR (data not shown). Phenotypes of csld1-1 and csld1-2 homozygous mutants, as well as offspring from their reciprocal F1 hybrids, were indistinguishable, thus demonstrating a causal role for the dysfunctional CslD1 gene. Mutant plants showed overall reduced growth, narrow leaves, and had a rough leaf texture caused by 9 warty protrusions from the mature leaf epidermis. Genotypic analysis of over 200 individuals from segregating families showed a 100% correspondence between this phenotype and homozygosity for the csld1-1 mutation (data not shown). Mendelian segregation ratios were typical of a recessive mutation. The five other transposon insertions in CslD1 (csld1-3 through csld1-7, courtesy of Pioneer Hi-Bred Int.) also resulted in mutant phenotypes that included narrower organs and wart-like clusters of epidermal cells, regardless of heterologous genetic background.

Phenotype of csld1 Mutants: Narrow Organs and Wart-Like Cell Clusters
Null mutants of csld1 showed a striking phenotype that included narrow leaves, reduced stature, and highly textured leaf blades . Light microscopy showed that the visibly rough texture of csld1 mutant leaves was due to irregular swelling by groups of epidermal cells that formed wart-like cell clusters (Fig. 2). Some epidermal cells expanded 75-fold in volume, and were generally arranged in linear profiles along the longitudinal axis of the leaf midrib and blade ( Fig. 2A), but not on leaf sheaths or stalks (data not shown). The Groups of cells that formed so-called warts were interspersed with normal-appearing regions of leaf epidermis along the entire length of the leaf blade. Swollen cells remained filled with fluid until the onset of leaf senescence (Fig. 2C). Warts were present on both surfaces of mutant leaf blades, but were larger and more abundant on the abaxial face (Fig. 2D). These malformed cells consistently lacked chloroplasts (Fig. 2C, E), indicating an epidermal origin, as confirmed in serial cross sections of mutant leaves (Fig. 2F, G). These ballooned epidermal cells in csld1 mutants often had diameters well over 100 µm, at least five-, and sometimes 20-fold greater than epidermal pavement cells of non-mutant plants (Fig. 2G). Both SEM and optical microscopy of fresh, intact leaves showed that lesions continued to expand throughout leaf maturation and that the largest clusters included swollen cells that had collapsed. In other instances cells remained intact, even in lesions greater than 300 µm across (Fig. 2C).
Potentially-analogous, wart-like epidermal swellings were described by Burton et al., (2000) in a VIGS gene silencing experiment in tobacco. Although a CesA gene was targeted, the highly-similar CslD genes may also have been silenced. The transgenic tobacco had wart-like lesions on the abaxial side of leaves and an overall phenotype strikingly similar to the maize csld1 mutants (Burton et al., 2000). This commonality would be consistent with some degree of repression of the Zm-CslD1 ortholog in the tobacco experiment. Alternatively, if observed lesions did result from down-regulation of CesA genes alone, then this would support a role for CSLD proteins in cellulose biosynthesis (Doblin et al., 2001;Park et al., 2011).

Plant Dry Weight and Organ Width are Reduced in csld1 Mutants
Total growth and organ size were reduced in homozygous csld1 mutants (Fig. 3) even though overall plant architecture, leaf number, and flowering time, were similar for csld1 mutant and non-mutant siblings under field and greenhouse conditions (data not shown). At maturity, the mean height of mutant plants (to auricle of the uppermost leaf) was only 11% less (p < 0.001), whereas dry weight decreased by a more pronounced 44% (p < 0.0025) (Fig. 3A).
Proportional reductions in dry weight were also evident for all organs examined, including ears, tassels, stalks, roots, and leaves (data not shown). Organ width decreased 35% (p< 0.0003) for mature leaf blades of csld1 mutant plants, but length was only 10% less (p < 0.0003) (Fig. 3B).
This narrow-leaf phenotype was proportional in all leaves examined, indicating a consistent defect in lateral development rather than an ontological effect at specific leaf positions (Fig. 3B).
Organ width was also reduced in stalks from csld1 mutants (Supplemental Figure 2). The cross-sectional areas of csld1 stalks were an average of 24% less (p < 0.025). The narrow-organ phenotype extended to cobs and tassels as well (data not shown).

Cell numbers are Reduced in Mutants and are More Pronounced than Changes in Cell
Size.
Both cell division and cell expansion contribute to the final shape of an organ; yet it has been traditionally difficult to separate causative factors during maize leaf morphogenesis. Given the reduced leaf width in csld1 mutants, we asked whether the mutant leaf morphology was due to decreased cell number, indicating CSLD1 acts directly or indirectly on cell division, or whether the primary cause of narrower leaves was the altered cell width. To answer this question, we quantified cell numbers and sizes in mutant and non-mutant leaf blade epidermis ( Fig. 4). Examination of impressions from non-warty areas of mutant and non-mutant leaves showed that the length of epidermal cells was not detectably different, and the width of cells from the csld1 mutant was 17% greater (p < 0.0003) (Fig. 4). Interestingly, the most pronounced difference between mutant and non-mutant was that of total cell numbers, and thus divisions, across the leaf blade, which was 47% less for csld1 mutants (Fig. 4) based on calculations from cell-level (Fig. 4) and leaf-level analyses (Fig. 3).
Leaf cross sections were examined to ascertain whether similar abnormalities were present in non-epidermal cells of csld1 mutant leaves. Leaf thickness was consistently increased in mutant leaf blades, and vascular bundles were closer together (Supplemental Figure 3A).
Fully-expanded csld1 mutant leaf blades were 40% (p < 0.004) thicker than non-mutant leaf blades (Supplemental Figure 3B). The extent of non-epidermal contributions to total leaf thickness was 43% (p < 0.002) greater for csld1 mutant leaf blades, and thus proportional to those of epidermis (Supplemental Figure 3B). Increases in cell size were variable but most often apparent in mesophyll-sheath cells. Overall organization of the mesophyll sheaths was less regular. The csd1 phenotype was therefore not limited to the epidermis alone. Additionally, mutant leaves showed a 12% (p < 0.03) increase in vascular-bundle density (Supplemental Figure 3C).

Levels of CslD1 mRNA are Greatest in Regions of Active Cell Division
In order to view the phenotypes of csld1 mutants in the context of where the wildtype gene is expressed, levels of CslD1 mRNA were measured via quantitative RT-PCR across diverse tissues and stages of development (Fig. 5). The CslD1 transcript levels were greatest in young, pre-emergent leaves (inside the whorl) and still relatively abundant in young primary root tips and bases of more mature leaves (Fig. 5A). These mRNAs were also detected in basal zones of leaves by other recent studies of maize leaf development (Li et al., 2010). To more clearly define the pattern of transcript accumulation during leaf development, staged samples of young to mature leaves were analyzed. The CslD1 mRNA levels were highest in tissues with actively dividing cells, and maximal in shoots 6 days after germination (Fig. 5B). Later in development, single-leaf analyses showed that levels of CslD1 mRNA were greatest in basal portions of blades from expanding leaves, 15-to-25 cm long. In non-dividing, fully-expanded leaves, CslD1 mRNA had dropped below detectable levels and remained so in the fully-differentiated portions of leaves (Fig. 5B). Notably, wart formation and expansion continues in leaves through maturity, even within areas where CslD1 is no longer expressed.

Cell Wall Composition Was Similar, but Thin, Dense Walls Were More Abundant in csld1
Mutants.
To further characterize differences between mutant and non-mutant tissues, highresolution X-ray micro computed tomography (micro-CT) was used to analyze hand-cut sections of stalks, since this tissue was found amenable to the X-ray approach. Three-dimensional reconstructions revealed a shift in average wall thickness towards thinner walls in the csld1 mutant (Fig. 6). Additionally, this method allowed a comparison of overall wall density, which was greater for stalks from mutant plants, regardless of whether the pith and vascular-rich rind were examined together or separately (Fig. 6).
To determine whether these changes were associated with alterations to cell wall polysaccharide composition, we examined cell walls from mature leaf blades as well as isolated leaf epidermis. No significant differences in alcohol-insoluble cell wall composition were detected between csld1 mutants and non-mutants for either cellulose or sugar subunits of noncellulosic constituents (Table 1). Cell walls from epidermal cells of both plant types revealed distinctive composition relative to samples from whole-leaf blades. Specifically, epidermal cell walls had less glucose, rhamnose, galactose, and galacturonic acid, but more xylose, compared to whole-leaf blades (Table 1).

Multiple Cell Division Defects are Evident in csld1 Epidermis
Dark-field images of epidermal peels from mature leaf blades showed that the normal ordering of epidermal cells was disrupted in csld1 mutants, and revealed frequent anomalies in recently-formed cross walls. The latter included stubs of incomplete cell walls, particularly along the longitudinal axis of leaves (Supplemental Figure 3). Other alterations of cell shape and misaligned cell walls were also consistent with causal defects in cell division, and these persisted in mature leaves.
To determine the developmental stage at which these cell abnormalities could first be detected, 5-10 cm immature leaves (sites of maximal CslD1 expression) were sampled from csld1-mutant and non-mutant plants, stained with propidium iodide, and examined by confocal microscopy (Fig. 7). Disrupted cell files and misshapen cells were evident in images of abaxial epidermal cells from the pre-and post-differentiation zones (as defined by Mitkovski and Sylvester [2003] and determined by the absence and presence of stomata, respectively) ( Fig. 7A, B). Cell wall stubs were frequently observed in epidermal cells of both pre-and postdifferentiation zones (Fig. 7). Again, these were typically oriented along the longitudinal axis ( Fig. 7). While many epidermal cells at these stages were approximately twice the normal width (as if they had failed to undergo a single longitudinal division), nearly all cells of csld1 mutants were at least some degree wider than those of non-mutants ( Fig. 7 A, B).
A greater cell width also remained evident later in development, even for otherwise normal-looking cells outside wart-like lesions (refer back to Fig. 4). Also, in multiple instances, files of atypically large epidermal cells were bounded at proximal and distal ends by cells that appeared to have mis-divided. Cell wall stubs were visible in each, and faced the large files.
These files of wide cells were located at positions normally occupied by two distinct files of smaller cells (Fig. 7C). This pattern indicated either failure of consecutive neighboring cells to divide, or more likely, a clonal file derived from a single cell that failed to complete division. In addition, three-dimensional imaging with confocal microscopy identified a large number of gaps and holes in cell walls that would otherwise have appeared to be complete when viewed by standard imaging techniques (Fig. 7D). The actual number of incomplete walls may thus be underestimated when non-confocal approaches are used.
Because cells with large or multiple nuclei are commonly observed in cell-divisiondefective mutants (Smith et al., 1996;Lukowitz et al., 1996;Spitzer et al., 2006), propidiumiodide stained nuclei were examined in basal regions of immature leaf blades from csld1 mutants ( Fig. 8A). Compared to non-mutant epidermis of the same stage (pre-differentiation zone of 5-10 cm leaves), csld1 mutant cells generally had larger nuclei (Fig. 8A). While significant variation in nuclear size was observed even in non-mutant tissue, the range in csld1 mutants was much greater, with cells containing nuclei ranging from one-, to approximately four-fold the two-dimensional area of normal nuclei (Fig. 8A). Large nuclear size was consistently correlated with large cell size (Fig. 8A). Flow cytometry of nuclei from basal regions of immature leaves showed significantly more (p < 0.05) 4N nuclei in csld1 mutants compared to non-mutants ( Fig.   8B). Integration of mutant phenotypes from maize, rice, and Arabidopsis with the CSLD phylogeny reveals ancient functional divergence and highly-conserved developmental roles for sub-groups within the CSLD family (Fig. 1A). Although double-and triple-mutant analyses have demonstrated some overlapping function for CslD genes in Arabidopsis (Yin et al., 2011), each single-gene mutation described thus far has a distinctive phenotype (except for a putative pseudogene At-CslD6) (Favery et al., 2001;Wang et al., 2001;Bernal et al., 2007Bernal et al., , 2008Kim et al., 2007;Li et al., 2009;Penning et al., 2009;Hu et al., 2010;Wu et al., 2010). When we overlay data on mutant phenotypes onto a phylogentic tree, CSLD clades correspond to distinct classes of phenotypes; (i) root-hair-defective, (ii) male-transmission-defective, or (iii) reducedgrowth (Fig. 1A).

CSLDs Acquired Divergent Roles Early in Plant Evolution
There are several notable aspects of this pattern. First, the observation that each CslD mutant has a visible phenotype indicates limited functional redundancy within the gene family.
Second, the associations shown in Figure 1A persist despite fundamental structural differences between the Type I primary cell wall of Arabidopsis and the Type II cell wall of rice and maize (Harris and Hartley, 1980;Carpita and Gibeaut, 1993;Carpita, 1996;Carpita et al., 2001). Roles of CslD genes thus transcend the major differences in non-cellulosic cell wall constituents of diverse plant species, and the primary developmental functions of individual CslD genes appear to have been maintained. Third, previous results suggested specific contributions by CSLD proteins in tip-growing cells (Bernal et al., 2008), however, disruptions in At-CslD5, Os-CslD4, and Zm-CslD1 led to reduced overall plant growth without visibly altering classic tip-growing cells (Fig. 1A, Bernal et al., 2007;Li et al., 2009;Hu et al., 2010;Wu et al., 2010;Yin et a., 2011). Broader developmental functions are thus indicated for these genes. Among the reducedgrowth phenotype clade, that of maize csld1 is unique in its production of wart-like lesions. This difference might lie in greater growth rate and expansion of maize leaves. In other respects, however, commonalities between the reduced-growth phenotypes suggest a shared function for this sub-group of CSLDs.

Reduced Cell Division Is a Central Effect of the csld1 Mutation
The combination of reduced leaf width, together with overly expanded cells is intriguing Expansion of the epidermal warts appears to be largely a secondary effect of the csld1 mutation, because these cells continue to expand in leaves after CslD1 mRNAs would have dropped below detectable levels (Fig. 5).
While results are compatible with a central defect in rate or total number of cell divisions, diverse, indirect effects may also contribute to the observed phenotype. Narrow leaves of csld1 mutants, for example ( Fig. 3), could exacerbate reductions in plant dry weight by decreasing total-plant photosynthetic capacity. The smaller csld1 root system (Fig. 3) could further reduce growth. Also, the epidermis may have a prominent physical role in organ expansion and meristem geometry (Green, 1980;Moulia, 2000) providing additional potential for secondary or tertiary effects of the csld1 mutations.
Cell expansion is clearly also altered by direct and/or indirect effects of the mutation.
Much of the cellular over-expansion is likely secondary, since compensatory expansion is a https://plantphysiol.org Downloaded on February 12, 2021. -Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved. common response to decreases in cell number (Reynolds et al., 1998;Mitkovski and Sylvester, 2003;Beemster et al., 2003;Horiguchi and Tsukaya, 2011). However, the csld1 mutations could theoretically affect more than cross-wall deposition during division and subsequent, compensatory expansion. If properties of mutant walls are altered, then these too could contribute to expansion-based aspects of the phenotypes (discussed further below).

Cell Wall Thickness, but Not Composition Are Altered in the csld1 Mutant
Although the overall quantity of cell wall material is reduced by 45% in the csld1 mutant plants, cell wall composition is unaffected in either whole organs or epidermis alone (Table 1).
Several possibilities could account for the lack of detectable difference. First, the CSLD1 polysaccharide product may normally be present in only small amounts, particularly if synthesis is restricted to a defined period during cytokinesis or new cell wall formation. If so, then cell wall contributions from CSLD1 could be masked by more abundant polymers. As noted above, an early developmental role for small amounts of CSLD product would be consistent with the maximal expression of CslD1 in very young leaves and basal portions of fast-growing blades ( Fig. 5), since these are zones of active cell division (Sylvester et al., 1990;Freeling, 1992;Sylvester, 2000). One permutation of this suggestion is that CSLD1 might aid a specialized, directed deposition of early-arriving polymers to the growing cell plate during division. The CSLD1 product itself could thus be limited to a narrow point in time.
Another possibility is that the CSLD1 enzyme might synthesize a limiting constituent of cell walls, such that a drop in levels of the CSLD1 product would result in similar decreases for other cell-wall polymers. In this way, csld1 mutants could produce less total cell wall without changing relative proportions of individual cell-wall components. This suggestion is consistent with the reduction in total dry matter. Still another, compatible scenario, is that the polysaccharide product of CSLD1 may not differ in a way immediately detectable by current analyses, but may nonetheless have functionally altered properties and/or capacities for interaction that affect later cell-wall behavior. All three of the above possibilities remain consistent with a temporally-limited contribution of a known wall component.
Cell wall analysis by high-resolution, X-ray, micro CT ( of csld1 mutants (tissues found most amenable to the X-ray, micro CT approach) show slight, but consistent increases in wall density and a greater overall abundance of thin-walled areas, regardless of position in rind or pith (Fig. 6). These findings indicate an altered cell-wall or cellular architecture that is apparently independent of changes in wall composition. One, albeit speculative possibility is that the changes in cell wall density may reflect a relative increase in cellulose crystallinity, with less amorphous cellulose being present in the cell walls of the csld1 mutant.

Role of CslD1 in division and expansion
Cell-level effects of csld1 mutations highlight differences between global and local responses that lead to narrow leaves and overly expanded, warty clusters, respectively (Fig. 9). In whole leaves, cell divisions are reduced markedly (almost 50%), while expansion is increased slightly (about 20%) (see also Figs. 3,4). In clusters of abnormal cells, however, division anomalies are more frequent and expansion effects more extreme. One reason for this distinction is likely a contrast in mechanisms by which cell expansion responds to altered cell number (Horiguchi and Tsukaya, 2011). At the whole-leaf level, classical compensatory expansion is widely recognized as a means of adjusting organ shape via cell enlargement (Reynolds et al., 1998;Mitkovski and Sylvester, 2003;Beemster et al., 2003;Tsukaya, 2008;Horiguchi and Tsukaya, 2011), whereas local responses to changes in cell number may differ and mechanisms remain unclear (Horiguchi and Tsukaya, 2011). Primary effects of CslD1 disruption clearly include cell division, and although many aspects of expansion may be secondary, some could also be direct.

Warts Are Distinctive, Informative Features of the csld1 Mutant
In warts, swelling epidermal cells (Fig. 2) trace to cell-division defects early in leaf development (Fig. 7). Assuming CSLD1 is a wall-synthesizing enzyme, these cellular observations indicate that loss of CSLD1 activity disrupts the synthesis of early cell wall components, and this in turn alters division. The narrow organs result from a decreased number of long-axis cell divisions throughout the plant (discussed above). Wart formation also derives from disrupted divisions, but the locally-extreme cell size may involve additional effects of csld1 on wall expansion. Two possible scenarios for epidermal wart formation are presented below.
Hypothesis 1 is based on a role for CSLD1 in new cell wall formation during cytokinesis, and posits that epidermal warts would be initiated when rapid elongation by leaves outpaces capacity of dividing cells to form normal cross-walls. The first set of disrupted cross-walls would lead to misplacement of the next, and resulting clusters of irregular cells would be prone to anomalous expansion. Extent and irregularity of expansion would be exacerbated by asymmetric support from surrounding cells. Hypothesis 2 includes basic tenets of Hypothesis 1, but also suggests that csld1 mutants may have broader alterations to cell wall properties that The sequence of events in hypothesis 1 begins with a central role for the CSLD1 polysaccharide product (likely cellulose) in formation of new cross-walls. In csld1 mutants, this cross-wall formation is implicated in reducing the number of long-axis cell divisions and thus organ width. In leaf blades especially, a CSLD1-based limitation to new wall formation could have prominent effects when the demands of rapid growth outpace the compromised capacity for cross-wall production. Resulting anomalies in csld1 epidermis include instances of non-divided cells and defective cross-walls. The latter could arise from either partial formation of these walls and/or their expansion before full completion. Resulting cells have non-existent or defective cross walls with central holes or cell wall stubs (as observed). A single cell division altered at the base of a growing leaf blade could lead to other disruptions by altering the timing and placement of the subsequent cross-walls. The abnormal divisions would then lead to the uneven and disrupted cell files, as well as large and misshapen cells (Fig. 7A, B). These cells would lack the support otherwise afforded a normal, highly-ordered, brick-like pattern of epidermal cells, and would be more prone to excessive expansion during turgor-driven growth. In addition, hypothesis 1 predicts that the warty-ness of csld1 phenotypes will be more extreme under conditions where growth is most rapid and cross walls cannot form correctly. Consistent with this prediction, phenotypes vary in plants grown under field-versus greenhouse-conditions (data not shown).
Another possibility is that csld1 alters cell-wall properties that affect expansion and these in turn cause cell division defects that lead to wart formation (Hypothesis 2). Such changes could have functional significance even if generated by trace amounts of materials contributed only during a brief period early in cell development. Although altered cell walls (missing a CSLD1 product) could be "weaker," they might also be more responsive to effectors of cell expansion such as endogenous signals or physical aspects of turgor pressure. Larger or irregularly-shaped cells could result, and this could compromise their capacity to form accurately-positioned or complete cross walls. The resulting series of anomalous, failed or disrupted divisions would produce warty cell clusters. The suggestion that csld1 cells expand more readily would be consistent with the somewhat larger size of epidermal cells overall. Wall properties might be altered in subtle ways throughout development or impart a threshold component to ballooning cells of warts.

The CslD1 Gene Is Essential for Specific Aspects of Cell Division
Primary effects of the csld1 mutation on cell division can be clearly discerned in the narrow leaves and fine stems of the csld1 phenotype. Although the overly-expanded warty lesions exhibit cascades of anomalous secondary events, the majority of cells in leaves and stems are well-ordered. Reduced cell numbers and cell divisions in these organs are consistent with a global response to the csld1 mutation. In particular, long-axis divisions are decreased, and proportional to reductions in leaf width, other organ diameters, and plant dry weight. Although narrowness of leaf blades could theoretically be exacerbated by secondary effects of warts, the same cannot be said of the other narrow organs of csld1 plants. As a cell-wall biosynthetic gene,  2003;Cnops et al., 2004;Hu et al., 2006;Szécsi et al., 2006;Horiguchi et al., 2011;Horiguchi and Tsukaya, 2011), and a primary role is suggested here for CslD1 in cell division.
Deposition and integrity of new cross-walls are clearly impaired in dividing cells of csld1 mutants. Long-axis divisions are most markedly affected, with new walls having large central openings or appearing as incomplete stubs. Collective data reveal a new dimension to functions of the CslD gene sub-family, since these were previously considered largely related to tip growth (Favery et al., 2001;Wang et al., 2001;Kim et al., 2007;Bernal et al., 2008;Penning et al., 2009;Park et al., 2011). However, both tip growth and cross-wall formation during division share a common, directional aspect to the deposition of new wall material. Bednarek and coworkers (Bednarek and Flabel, 2002) previously suggested that such mechanisms could involve similar components. Work here is consistent with that possibility, as well as contributions by CslD genes to targeted wall formation in tip-growing and dividing cells.
The csld1 mutation also disrupts cytokinesis and/or the cell cycle. Larger nuclei were observed by confocal imaging in cells of the pre-differentiation zone of immature csld1 leaves ( Fig. 8A). Similar, large-nucleate cells have been reported in studies of cell-division-defective mutants (Lukowitz et al., 1996). Here, flow cytometry also showed a small, but significant increase in endoreduplication (Fig. 8B). Even with 50% of epidermal cells undergoing endoreduplication, these would comprise a relatively small portion of nuclei from whole-leaf tissue. Whether the increase in endoreduplication and larger-sized nuclei reflects a response to larger cell size or the arrest of the cell cycle after DNA replication, but before nuclear division, remains unclear. A more rapid entry of dividing cells into an endoreduplicative state is typically observed where cell size is increasing as a compensatory response to limited cell number in a developing organ (Beemster et al., 2005;Ferjani et al., 2007;Horiguchi and Tsukaya, 2011).
The delay of cells at the G2 phase implies a disruption of the normal cell cycle due to a lack of CSLD1 activity.

CONCLUSION
The significance of collective findings is threefold. 1) A previously-unrecognized role in cell division is demonstrated here for a cell-wall biosynthetic gene. 2) CslD1 is also found here to be central for effective formation and integrity of new cross-walls during cell division. 3) Collective roles for the CslD gene subfamily are shown to include not only tip-growth by cells, but also probable contributions to other aspects of directional wall deposition. Additionally, csld1 maize mutants, besides helping our understanding of the specific functions of a clearly important cell-wall synthesizing enzyme, will provide valuable tools for dissecting the complex and interconnected processes of cell division and cell expansion. Protein localization efforts are underway and should further clarify the specific role of Cellulose Synthase-Like D1 in early cell wall formation.

Phylogenetic Analyses
Protein sequences predicted from full length cDNAs for each of the CslD genes from rice, maize, and Arabidopsis were used to create a neighbor-joining tree using

Identification of csld1 Mutants
The UniformMu population was screened using PCR-based assays to identify Mu transposon inserts in Zm-CslD1 as per Penning et al., (2009). Close to 15,000 UniformMu lines were screened using a series of pooled DNA samples, which were forerunners of the sequenceindexed materials currently available at MaizeGDB (maizegdb.org; UniformMu.UFgenome.org). For PCR screening, CslD1-specific primers (AGTTCGTGCACTACACCGTGCACATCC and TGCTACCTGTAAGGACTGAGGATGGCCTG) were used along with the Mu-specific primer TIR6 (AGAGAAGCCAACGCCAWCGCCTCYATTTCGTC). Resulting products were separated on 1% agarose gels, blotted onto nylon membranes, and probed with a CslD1-specific PCR product.
Positive probe-binding samples were identified at X45:Y4 of the UniformMu Reverse Genetics Grid 6 (of 8 total). Seeds from the UniformMu family corresponding to these coordinates (04S-1130-27) were grown, and PCR-genotyped to identify individuals homozygous for an insertion in CslD1. The csld1-1 allele was identified from this family, and a csld1-1 line was established after three generations of successive backcrosses to the W22 inbred.

Overall Phenotypic Analyses and Size Measurements
Plant height was measured from soil level to auricles at the base of the uppermost leaf blades for 55 field-grown, mature plants (25 mutant, 30 non-mutant). These same 55 plants were used to measure width (at the widest point) and length of leaves at positions three, four, and five (relative to the apex). For dry weight measurements, whole-plant samples (including released root mass) were collected 3 days after ear harvest and did not include mature ears. Samples were weighed after drying for 4 weeks at 38°C. Below-and above-ground dry weights were determined by separating root masses from the aerial portions of these plants.
Paraplast wax chips (Fisher Cat # 23-021-399) (1 g wax/mL CitriSolv) were added to the 100% CitriSolv and incubated overnight at 25°C. Additional wax was added, followed by a 2-hr incubation at 42°C. Samples were transferred to 60°C where wax was poured-off and replaced eight times over 4 days before samples were allowed to harden in molds. Sections (10 µm, cut with a Leitz 1512 microtome) were de-waxed with three, 5-min incubations in xylene (Fisher Lot # 083423), then washed twice in 100% EtOH (5 min each), and once in 95% EtOH (3 min).
Slides were dried and examined under a Olympus BH2 light microscope.

Cell Volume Estimates
The extent of maximal expansion was estimated for ballooning epidermal cells of the

Epidermal Impressions and Non-warty Cell Size Determination
Fresh samples from mature leaves of greenhouse-grown plants were cut into 1-cm 2 pieces and firmly pressed into Superglue on glass slides. Glue was allowed to dry completely before leaf tissue was removed, leaving detailed epidermal impressions. These were imaged under a light microscope (Olympus BH2) with an RT SPOT camera (Diagnostic Instruments). Average cell length and width were determined by quantifying the total number of cells in a given area (1.88 x 1.40 mm). Longitudinal and lateral transects were used that did not include warty clusters. Resulting RNA was treated with DNase-1 (Ambion Cat # AM1906), and quantified using a BioRad SmartSpec 3000. The cDNA was synthesized using SuperScript One-Step kit and protocol (Invitrogen Cat # 10928-042).

High-resolution X-ray Computerized Tomography Analysis
Field-grown plants were sampled three days after harvesting ears and dried at 38°C for three weeks. Sections (~0.5 cm) from the middle of the second internode of the conditioned stems (~9% moisture content) were cut using a small band saw and scanned using a Scanco Medical Ag uCT35 instrument (Brüttisellen, Switzerland). Initial measurements were taken on whole-stem sections at 10-micron resolution. Regions including pith and rind (3 x 4 mm) were hand-cut from the edge of these sections and scanned at 3.5-uM resolution over a 0.88 mm high region to quantify cell wall and air space sizes. The 232 slices from each scan were reconstructed into three dimensional images and contoured over whole stems for volumetric analyses. All scans were conducted with integration times of 600 microseconds and averaging two times. Both a fixed, common threshold and an adaptive threshold were used to segment cell wall from airspace and volumetric analyses were calculated with an algorithm developed for trabeacular bone (Hildebrand and Ruegsegger, 1997). For rind-only analyses, hand-drawn contours were used to isolate the vascular bundle-rich region along the edge of the stem prior to 3D reconstruction. Samples from leaves and epidermal peels were ground in liquid nitrogen along with 200 µL of extraction buffer (50mM Tris-Cl with 1% SDS at pH 7.2). Homogenate was transferred to 14-mL polypropylene, round-bottom tubes (Falcon product # 352059) along with 9 mL extraction buffer, incubated for 15 minutes at 80°C, and centrifuged at 3500 rpm for 5 min (~2,000 x g) in a swinging-bucket rotor centrifuge (ThermoForma 1LGP). Supernatant was removed with an aspirator, and pellets (water-insoluble cell-wall fraction) were washed, resuspended, and re-pelleted three times in about 10 mL 80°C water. The same process was repeated three times with 50% EtOH at 80°C, followed by three washes with 80°C water.

Cell Wall Composition Analysis
Samples were transferred to 1.5-mL Eppendorf tubes, alcohol-insoluble cell wall fractions were pelleted and dried, and non-cellulosic sugar composition was analyzed by the Complex Carbohydrate Research Center (University of Georgia; Athens, GA).
For cellulose content, the alcohol-insoluble cell wall fractions from whole-leaf samples were isolated in the same way, dried for 30 h at 60°C, transferred to 14-mL polypropylene, round-bottom tubes, and weighed. For each sample, approximately 50 mg of cell wall isolate was used, to which 3 mL 80% aqueous acetic acid and 300 µL 70% nitric acid were added.
Tubes were incubated in an oil bath at 110°C and 120°C for 20 min each, to hydrolyze hemicellulose and lignin (from Sun et al., 2004). Samples were cooled, 1.8 mL distilled water was added, tubes were centrifuged for 5 min (~2,000 x g), and supernatant was removed with an aspirator. Celluose was rinsed thoroughly with water (3 times) and 95% EtOH (3 times), and dried for 30 h at 60°C. Samples were weighed and compared for cellulose content as a fraction of alcohol-insoluble cell wall isolate.

Propidium Iodide Staining
Immature leaves (10 to 15 cm) were dissected from whorls of csld1 mutant and nonmutant plants. The basal portions (2 cm) of these leaves were immediately submerged in a solution of 0.1 mg/mL propidium iodide, and allowed to absorb the dye for 5 min at 25°C.
Samples were then rinsed thoroughly in water to remove excess stain and flattened on a glass slide. The abaxial epidermis was imaged using a Zeiss confocal microscope. For visualization of nuclei, the same process was followed, but leaf samples were first fixed in FAA (10% formaldehyde [Fisher Lot # 992720], 5% acetic acid, 50% EtOH), before staining with propidium iodide.

Accession Numbers
Accesion numbers for each of the gene sequences referred to in this work are: At-      1.40mm). Axes used for analyses of epidermal cell size from mutant leaves were done with transects that did not cross cells in the ballooning protrusions. Approximate cell numbers across and along leaf blades were estimated by comparing cell size estimates with average leaf length and width measurements (see Figure 3). These analyses did not include the midrib (0.5 cm was subtracted from total leaf width). Mean width of epidermal cells was 17% greater in csld1 mutants, whereas mean cell numbers were 48% less across the lateral axis and 16% less along the longitudinal axis. Values are marked with * where they differ significantly from those of non-mutant plants (p < 0.05) .        , and did not include mature ears or fine roots. Total plant dry weight was 41.9 g (SE, 4.0; N = 4) for mutants and 76.5 g (SE, 3.2; N = 4) for non-mutants. Below-ground dry weight was 8.7 g (SE, 0.2; N = 4) for mutants and 17.1 g (SE, 1.1; N = 4) for non-mutants. Above-and below-ground dry weights were reduced by similar amounts (av. 44% and 49%, respectively). Standard errors are indicated with vertical bars, and are marked by * wherever values for mutant plants differed significantly from those of non-mutants (p < 0.05) . B, Leaf blade length and width (at widest point) were quantified for leaf positions 3 through 5 (as diagramed) for field-grown mutant and non-mutant plants. Non-mutant plants included both wildtype and heterozygous individuals from segregating progeny after 3 backcrosses into the W22 inbred. Imaged blades were from leaves in position #3 from the apex. The left-most portion of each graph shows combined data from all leaf positions measured. The leaf blade width-to-length ratio was 27% less for mutant plants. Blade width and length were reduced 35% and 10% respectively, relative to those of non-mutant plants. Data were similar for csld1-2 (not shown), and visual appraisals of the five other csld1 mutants. Note: Leaf photos were from greenhouse-grown plants, whereas quantifications were from field-grown plants. Significantly different from non-mutant (p < 0.05) indicated by *. Epidermal impressions from abaxial surfaces of fully-expanded leaf blades from mature, greenhouse-grown plants were used to estimate cell dimensions and cell numbers along leaf axes. Non-mutant plants included both wildtype and heterozygous individuals from segregating progeny after 3 back-crosses into the W22 inbred. Cell numbers were quantified (each designated by a tic mark on impressions such as those shown to the right) along longitudinal and lateral axes of defined length. Mean cellular dimensions were determined by dividing number of cells along a given axis by the length of that axis. (mutant N=14; non-mutant N=10; longitudinal axis 1.88mm; lateral axis 1.40mm). Axes used for analyses of epidermal cell size from mutant leaves were done with transects that did not cross cells in the ballooning protrusions. Approximate cell numbers across and along leaf blades were estimated by comparing cell size estimates with average leaf length and width measurements (see Figure 3). These analyses did not include the midrib (0.5 cm was subtracted from total leaf width). Mean width of epidermal cells was 17% greater in csld1 mutants, whereas mean cell numbers were 48% less across the lateral axis and 16% less along the longitudinal axis. Values are marked with * where they differ significantly from those of non-mutant plants (p < 0.05) .  u t a n t M u t a n t Figure 6. High-resolution X-ray micro CT analysis of csld1 mutant and non-mutant stalks. Stem sections from the third internode of greenhouse-grown csld1 mutant and non-mutant plants were scanned at 3.5µ m resolution using high-resolution X-ray micro-CT. Sections for scanning (approximately 3 x 2 x 10 mm) were hand-cut from the edges of stems from mutant and non-mutant plants (four each). Non-mutant plants included both wildtype and heterozygous individuals from segregating progeny after 3 back-crosses into the W22 inbred. Three-dimensional analyses revealed significant differences in density of wall material (lower left) and distribution of wall thickness (right). Mutant stems had more dense, but generally thinner walls, even when analyses were limited to the rind only. Volumetric analyses were calculated with an algorithm developed for trabeacular bone (Hildebrand and Regsegger, 1997).